1. Field of the Invention
The present invention relates to a process for producing a silicon single crystal.
2. The Prior Art
As is known, the Czochralski method involves pulling a single crystal from a melt with the aid of a seed crystal, the melt being provided in a crucible. A single crystal obtained in this way (Czochralski=Cz single crystal) should have as few crystal defects as possible (as grown defects). This is because these defects can cause critical problems in the subsequent production of electronic components. The same is also true of a single crystal which is produced using the float zone method (=FZ single crystal). The FZ differs from a Cz single crystal, in part, by a normally substantially lower oxygen content.
It is known that in silicon single crystals, the formation of defects depends, in part, on the pulling rate and the temperature gradient at the interface of the growing single crystal and the melt. If the ratio V/G(r), while the crystal is being pulled, is above a critical constant c.sub.crit, with c.sub.crit =1.3*10.sup.-3 cm.sup.2 min.sup.-1 K.sup.-1, V being the pulling rate and G(r) the axial temperature gradient at the interface of the single crystal and the melt, excess vacancies are formed. They form during the crystal growth and, when the crystal is cooled, aggregate to form "microholes" or voids (approximately 50-100 nm). Depending on the preparation method with which they are detected, these defects are referred to as D defects, crystal originated particles (COP) or flow pattern (FP) defects.
The higher the density of these defects, the worse is the gate oxide integrity (GOI) in electronic components which are the end products of the subsequent processing of the single crystals. If V/G(r) is below c.sub.crit, an excess of Si interstitial atoms is formed, which agglomerate to form so-called L pits (Cz crystals) or A swirls (FZ crystals). These Si interstitial aggregations produce, as secondary defects, extended dislocation loops (several .mu.m) which are especially detrimental to component production. Further, L pits reduce the mechanical strength of the silicon, which can be observed by an increased likelihood in terms of slips during the component production process. None of the aforementioned defects are found if the ratio V/G and the constant c.sub.crit are equal when the single crystal is being pulled. Since the axial temperature gradient G(r) increases monotonically with increasing radial distance r from the center of the crystal to the edge of the single crystal, there are considerable technical pulling difficulties in setting this latter very desirable condition throughout the crystal volume. In crystals which are pulled approximately under this condition, a region at the center of the crystal with vacancy defects is almost always found. This is followed with radial symmetry by an outer region with L pits (A swirl). In Cz crystals, a narrow strip with oxidation-induced stacking faults (OSFS) is formed at the annular boundary between the two regions. In FZ crystals, an annular denuded zone is found instead of the OSF ring.
In order to avoid the particularly detrimental L pits, all industrially used Cz crystals have to date been pulled with excess vacancies, i.e. V/G(r)&gt;c.sub.crit over the entire crystal radius. Attempts have been made to keep the density of the vacancy defects as low as possible. It is known that the oxygen content has only a minor influence on the defect density, and therefore on the GOI quality (C. Hasenack et al., Proc. 173rd Meeting Electrochem. Soc., 447 (1988)).
It has long been known that with FZ crystals, both vacancy and Si interstitial defects can be simultaneously suppressed by a low level of nitrogen doping (approx. 10.sup.14 atoms cm.sup.-3). This gives an almost perfect GOI quality. This positive effect in terms of the GOI quality is, however, lost when doping with oxygen as well. This is inevitably the case with Cz crystals because of the use of a quartz crucible (W. v. Ammon et al., Proc. of the Satellite Symp. to ESSDERC 93, Grenoble, The Electrochem. Soc., Vol. 93-15, 36 (1993)). In the case of FZ crystals doped with oxygen and nitrogen it is admittedly true that the nitrogen doping can also establish an improvement in the so-called B+ mode failures in electrical stress tests. But this is of no importance since it is only the percentage of capacitors which reach intrinsic failure (C+ mode) which is of interest to component producers. In Cz crystals, the defect densities are at another order of magnitude compared to FZ. (FZ crystals are not comparable with Cz crystals because of the entirely different thermal history, substantially higher purity, more rapid pulling rate and entirely different process control). In Cz crystals, only nitrogen doping methods have to date been reported with the purpose of reducing the vacancy defect density and therefore increasing the GOI quality (JP-06,271,399 A). In terms of defect reduction/GOI improvement, however, no quantitative indications whatsoever are given.
The doping of Cz crystals with nitrogen accelerates the precipitation of the oxygen in the single crystal (R. S. Hockett, Appl. Phys. Lett. 48, 1986, p. 224). While the pulled single crystal is cooling, the oxygen starts to precipitate at even higher temperatures than otherwise is usual. This leads to larger precipitation seeds, which in turn during an oxidation treatment carried out subsequently on a semiconductor wafer obtained from the single crystal, produce stacking faults on the surface of the semiconductor wafer. Further, the large precipitates do not dissolve fast enough in the near-surface region of the wafer, during the component production process. Thus it becomes difficult to obtain a denuded zone with the depth specified by the component producer.